The explosion of interest in artificial intelligence (AI) has sent shockwaves across the entire business world. Whole sectors are falling over themselves to figure out the ramifications of investing in AI for the future — and the implications of failing to do so.
Much of the coverage in recent months has revolved around generative AI to produce content at speed. However, for those looking to tap into AI’s truly transformative potential — and not just one that can send out the odd email — it’s a whole different ball game. For this technology to have the kind of impact people are imagining, AI needs access to plenty of training data from which it can learn — and significant computing power.
The snag is, while both of these pillars are essential, they also provide a challenge in terms of storage and data management. That’s why it should come as no surprise that the cloud has become the go-to solution for storing and processing data for AI applications.
Cloud storage offers a flexible, scalable, and cost-efficient solution for handling vast amounts of data. As such, cloud storage and AI applications are a perfect fit. Together, they serve as the repository for training data that allows machine learning models to make predictions or decisions based on new inputs.
Imagine, for example, a bank developing an AI-powered fraud detection system. For it to work, the machine learning model needs a wealth of transaction data — such as the amounts of cash spent, what the money was spent on, and where the transactions occurred. This would allow the AI program to spot anomalies, flag suspicious activity, and learn to spot fraudulent activity.
On its own, collating all this data is a mammoth task. But that is nothing compared to the data volumes and processing power needed for the AI-enabled anti-fraud system to work.
One way around this is to use a third-party cloud storage provider which would enable the bank to store and analyze its transaction data without having to invest or maintain its own physical infrastructure.
As the project grows, the bank may choose to employ multiple clouds for its anti-fraud AI system to help keep a lid on costs and ensure it complies with financial regulations. As the project gains momentum, the bank may decide to use one cloud platform for running the training phase that allows the AI program to ‘learn’ from the data…and another during the ‘inference’ phase when it makes its fraud-busting predictions.
This multi-platform approach is common in the development of AI systems. But it does mean that companies need to keep moving data around between different cloud platforms — often provided by multiple cloud providers — to capitalize on the potential of AI.
But there’s a problem. Moving data around between different platforms may be subject to egress — or data transfer — fees. These are charges imposed by cloud service providers when data is transferred out of their networks.
The issue is, that although egress fees are relatively small, they can quickly add up — especially for organizations utilizing multiple cloud providers and transferring vast data volumes.
To minimize egress fees — which are sometimes viewed as a ‘tax’ on data transfers — some cloud providers are encouraging their customers to store data and train AI models solely within their cloud. Yes, this negates the need for egress fees. But this approach is not always viable. Nor is it necessarily best practice.
The obvious solution to the problem is to see egress fees abolished altogether so that data can move freely without financial constraints. A future free of these fees would empower companies to store and analyze data across multiple clouds enabling them to use the best available tools. without incurring extra costs.
Such a move would allow organizations to fully harness AI’s potential without concerns about rising expenses.
Once implemented, a zero egress fee cloud storage model would result in significant cost savings for organizations, freeing up resources for other essential business areas. It also eliminates the risk associated with relying on a single cloud provider, ensuring greater reliability and better protection against outages.
Perhaps most importantly, scrapping egress fees would fuel innovation. The flexibility offered by a multi-cloud architecture allows businesses to effortlessly select the most suitable provider for specific tasks. This would enable organizations to focus on experimentation and innovation, leveraging AI and other cutting-edge technologies without being hindered by costs or limitations.
There is little doubt that AI stands poised to revolutionize industries and society as a whole. But it cannot do so without cloud computing, storage, and the movement of vast amounts of data between platforms.
As already highlighted, egress fees aren’t just costly. They are a barrier to innovation. If we’re to create a brighter future for AI in the cloud, eliminating egress fees is essential. It’s part of the reason why in 2018, Cloudflare joined the likes of Azure, Google Cloud, Oracle, Alibaba Cloud, and others to found the Bandwidth Alliance to help customers save on egress fees.
By adopting such an approach, organizations are free to harness AI’s full potential without concerns about the costs associated with transferring data between clouds. It also represents a significant advancement in multi-cloud storage, laying the foundation for a more efficient, innovative, and promising future in AI and data management.
This article is part of a series on the latest trends and topics impacting today’s technology decision-makers.
This article was originally produced for Business Reporter
John Engates — @jengates
Field CTO, Cloudflare
After reading this article you will be able to understand:
Why egress fees are seen as a tax on data
A model for zero-egress object storage
The impact zero-egress can have on realizing AIs full potential